[vdr] Re: Driver or VDR Problem setting and jusageing cutmarks on Nexus 2.1

[Klaus Schmidinger <Klaus.Schmidinger@cadsoft.de>]

"Dr. Werner Fink" wrote:
> 
> On Thu, Jun 03, 2004 at 10:03:21PM +0200, Klaus Schmidinger wrote:
> > Juergen Sauer wrote:
> > >
> > > Am Donnerstag, 3. Juni 2004 18:29 schrieb Klaus Schmidinger:
> > >
> > > > > I tried 1.3.6|9 + 2.6.6|7pre2 tose are running so-la-la, except
> > > > > a problem with the fine justage of cut marks.
> > > > >
> > > > > vdr hangs and is quitted by the watchdog, if I try to move the
> > > > > cut-marks.
> > >
> > > > Try
> > >
> > > >   export LD_ASSUME_KERNEL=2.4.1
> > >
> > > > before running VDR.
> > >
> > > That's it. Works fine.
> > > What's the reason for the 'incomability' ?
> > > Is the ld system buggy
> >
> > The problem is that kernel 2.6 based systems apparently use the
> > NPTL ("Native Posix Thread Library"), which VDR doesn't seem to
> > work correctly with.
> >
> > Since I myself don't use kernel 2.6 nor the NPTL I can't do any
> > testing here. So somebody who uses these will need to look into
> > this and find out what's wrong. I've said this on several occasions,
> > but apparently there is nobody interested in doing so...
> 
> One problem could be signal handling.  For NPTL the signal masks
> should be set in such way that for the thread which should receive
> a signal unblocks this whereas the other should block it.  Before
> any pthread_create() a simple sigprocmask() is usefull to unblock
> or block signals in the main process, whereas within the threads
> the pthread_sigmask() should be used to protect the threads
> from catching signals.
> 
> Maybe it would help to modify vdr.c in such a way, that all
> handled signals will be unblock in main VDR whereas blocked
> in the threads.  Don't know if the patch in the attachment
> work, neverthelss you may give it a try.
> 
>        Werner
>
>                         Name: vdr-nptl.patch
>    vdr-nptl.patch       Type: Plain Text (text/plain)
>                  Description: vdr-nptl.patch

I'll gladly apply this patch if somebody with kernel 2.6 / NPTL can
confirm that it solves the problem.

Klaus